用改进的多智能体遗传算法求解旅行商问题

Solving TSP with improved multi-Agent genetic algorithm

多智能体遗传算法是基于智能体对环境感知与反作用的能力提出的一种新的函数优化方法,具有很快的收敛速度,尤其是在优化超高维函数时更显示出了它的优越性。针对这一特点对该算法进行了适当的改进,在邻域正交交叉算子中采用精英保留策略,在自学习算子中引入邻域正交交叉算子并采用小变异概率以加快收敛速度。求解TSP的实验结果显示,改进后算法的性能有了较大的提高。

Based on agent＇s capability of perceiving and reacting on environment, Multi-Agent Genetic Algorithm （MAGA） was proposed as a new method of function optimization. MAGA had a rapid convergence velocity especially when it optimized super-high dimensional functions. This algorithm was improved properly based on its characteristics： elitist reservation strategy was adopted in neighborhood orthogonal crossover operator, and neighborhood orthogonal crossover operator was introduced into self-learning operator and small mutation probability was adopted to quicken the convergence speed. The results of solving Traveling Salesman Problem （TSP） show that the performance of improved MAGA is enhanced greatly.